Medications are usually introduced into clinical practice following a demonstration of superiority over a placebo. But what clinicians, patients, and families really want to know is how treatments for a given condition compare to each other with respect to therapeutic benefits and safety. In many cases, the evidence base for discriminating among treatments is limited, thus leaving clinicians, patients, and families to choose treatments on the basis of anecdotal experience or, worse, under the influence of marketing pressures.
Objectively, there are reasons for this dearth of information. Proper comparison between treatments requires testing them directly in the same randomized trial. Attempts to draw comparisons across trials, for example, by contrasting separate placebo-controlled trials of one drug against placebo-controlled trials of another drug, can lead to invalid conclusions due to variable placebo response and experimental sensitivity. From a methodological point of view, trials of active treatments are more challenging than mere comparisons to placebo. The difference in outcome between active treatments is smaller than between an active drug and placebo, hence the need to enroll a large number of subjects to ensure adequate statistical power. In order to be able to interpret an eventual lack of statistically significant difference between treatments as evidence of similar effectiveness, a noninferiority design is required, which, however, also requires a large study group. The inclusion of a placebo arm is desirable as it helps to determine the assay sensitivity of the study and therefore to infer treatment effectiveness (1).
To these design complexities, one needs to add the challenge of choosing the most appropriate formulation and dosage for each treatment. Few studies actually meet all these standards, and too many suffer from a small study group or suboptimal dosing (use of either excessively low or high doses can occur), thus creating unbalanced and biased trials.
Exceptions fortunately exist, as shown by the report by Newcorn et al. (2), who directly compared a commonly used extended-release formulation of methylphenidate with atomoxetine within a large randomized, controlled clinical trial in children and adolescents with attention deficit hyperactivity disorder (ADHD). The study incorporated many desirable features, including randomized design, placebo arm, large number of subjects, and optimal formulations and dosing strategies. The results indicate that, while both drugs are better than placebo, methylphenidate is more effective in decreasing ADHD symptoms than atomoxetine. Methylphenidate had a large effect size (d=0.8) with a highly favorable number needed to treat (NNT) of 3. Atomoxetine had a medium effect size (d=0.6) with a higher, less favorable, but still very acceptable, NNT of 5.
The study conclusively confirms previous, methodologically less rigorous, reports of stimulant superiority over atomoxetine in the treatment of children and adolescents with ADHD, and it provides support to the current treatment guidelines (3, 4). It should be noted that the study was funded by the pharmaceutical company that markets atomoxetine and that the report is the result of a collaboration between academic and industry investigators. While the ultimate decision of which treatment modality (i.e., psychosocial, pharmacological, or combined) and, in the case of pharmacotherapy, which medication to use first rests with the clinician and the parent, there is now ample evidence that stimulants are the most effective treatment for decreasing symptoms of ADHD.
The study also documents differences and similarities in the safety profile of these medications. Sleep difficulties were more frequent with methylphenidate, and somnolence with atomoxetine. Both drugs decreased appetite, but loss of weight was greater with methylphenidate. Both increased diastolic blood pressure, but atomoxetine increased heart rate also. These data confirm that stimulants and atomoxetine, consistent with their adrenergic activity, induce cardiovascular changes, which are on average small but whose long-term impact deserves further investigation.
Like with any clinical trial, this study also has limitations. The duration of treatment was 6 weeks: enough to detect acute effects but too short for measuring the impact on level of functioning or on long-term safety issues, such as growth suppression (5). A potential bias was introduced by including responders to previous methylphenidate treatment, while excluding patients who had not responded or could not tolerate it. But the presence of a bias is not confirmed, given the finding of a greater response rate among methylphenidate-naive (64%) than among previously treated (51%) patients. Other limitations are the asymmetric crossover after the initial 6-week trial, as only the methylphenidate group was switched to atomoxetine, and the lack of placebo control in this phase of the study.
Despite these limitations, the data provide support for the presence of preferential treatment response: nearly half of the patients who had not responded to methylphenidate responded to atomoxetine, while, of the methylphenidate responders switched to atomoxetine, 18% showed an even greater response and 36% showed a worse response. Preferential response has also been observed among children with ADHD treated with different stimulant medications. A review of within-subject comparisons of methylphenidate and amphetamine found that while 41% of patients improved equally well with either medication, 44% responded better to one of them (6).
This preferential treatment response is likely to be genetically determined (7). As in other areas of medicine, in the treatment of ADHD there is a need for investigating individual variation in treatment response and elucidating the underlying mechanisms as a step toward personalization of care (8).